
% Author: Kazuma Takakura
% Date: 17 May 2025
% Time: 11:12:54

\begin{table}[h!]\footnotesize
  \centering
  \caption{Heterogeneity among Baseline Abilites (Math and CPCS)}
\label{tab:hetero_cpcs}
\scalebox{1}{
\begin{threeparttable}
\begin{tabular}{lccccccc}\toprule
 Dependent Variable & \multicolumn{2}{c}{Baseline^{b}} & Difference & Obs & N of clusters  \\\midrule\midrule
 Rapid math test score^{a} & Math Top 50\%  & CPCS Top 50\% & -0.240 & 55 & 21 \\ 
                      & & & ( 0.872 )  & &   \\ 
  & & CPCS Bottom 50\% & -0.566* & 67 & 24 \\ 
                      & & & ( 0.092 )  & &  \\ 
  & Math Bottom 50\%  & CPCS Top 50\% & 0.030 & 53 & 23 \\ 
                      & & & ( 0.942 )  & &  \\ 
  & & CPCS Bottom 50\% & -0.192 & 68 & 24 \\ 
                      & & & ( 0.480 )  & &  \\ 
 CPCS score^{a} & Math Top 50\%  & CPCS Top 50\% & 0.774*** & 52 & 20 \\ 
                      & & & ( 0.006 ) & &   \\ 
  & & CPCS Bottom 50\% & 0.358 & 65 & 24 \\ 
                      &       & & ( 0.154 )  & &  \\ 
  & Math Bottom 50\%  & CPCS Top 50\% & 0.279 & 53 & 23 \\ 
                      & & & ( 0.546 ) & &   \\ 
  & & CPCS Bottom 50\% & 0.331 & 66 & 24 \\ 
                      &       & & ( 0.420 ) & &   \\ 
\\\midrule
\end{tabular}
\begin{tablenotes}
\item (a) Dependent variables are standardized using the average and variance of the whole follow-up sample in the February 2022 survey. 
\item (b) Cutoffs are created based on whether their ability to perform each item at the baseline was higher or lower than the median. 
\item (c) Wild cluster bootstrap p-values are reported within parentheses. Clusters are schools at the baseline.
\item (d) $^*$ Significant at 10\% level; $^{**}$ significant at 5\% level; $^{***}$ significant at 1\% level. 
\end{tablenotes}
\end{threeparttable}
}
\label{tab:addlabel}%
\end{table}




